Journal of Constructional Steel Research 63 (2007) 751765

www.elsevier.com/locate/jcsr

Cyclic loading behavior of EBF links constructed of ASTM A992 steel

Taichiro Okazaki a, , Michael D. Engelhardt b
a Department of Civil Engineering, University of Minnesota, Minneapolis, MN, 55455-0116, USA
b Department of Civil, Architectural, and Environmental Engineering, University of Texas at Austin, Austin, TX 78712-0275, USA

Received 4 May 2006; accepted 8 August 2006

Abstract
Cyclic loading tests were conducted to study the behavior of link beams in steel eccentrically braced frames. A total of thirty-seven link
specimens were constructed from five different wide-flange sections, all of ASTM A992 steel, with link length varying from short shear yielding
links to long flexure yielding links. The occurrence of web fracture in shear yielding link specimens led to further study on the cause of these
fractures. Since the link web fracture appeared to be a phenomenon unique to modern rolled shapes, the potential role of material properties on
these fractures is discussed. Based on the test data, a change in the flange slenderness limit is proposed. The link overstrength factor of 1.5, as
assumed in the current U.S. code provisions, appears to be reasonable. The cyclic loading history used for testing was found to significantly affect
link performance. Test observations also suggest new techniques for link stiffener design and detailing for link-to-column connections.
c 2006 Elsevier Ltd. All rights reserved.

Keywords: Cyclic tests; Steel structures; Seismic design; Flange slenderness ratio; Loading history; Fracture; k-area; Eccentrically braced frame

1. Introduction
The design intent for a seismic-resistant steel Eccentrically
Braced Frame (EBF) is that inelastic action under strong
earthquake motion is restricted primarily to the links. Therefore,
the EBF design procedure prescribed in the 2005 AISC Seismic
Provisions for Structural Steel Buildings [1] relies on an
understanding of link behavior under severe cyclic loading.
The AISC Seismic Provisions contain U.S. building code rules
for detailing steel structures, including EBFs, for seismic
resistance. The current building code rules for EBFs in the AISC
Seismic Provisions, including link design, link rotation limits,
and link overstrength factors, were developed from rather
extensive experimental studies conducted almost exclusively
on wide-flange shapes of ASTM A36 steel [2]. However,
structural steel shapes most commonly used in the U.S. today
are produced according to the newer ASTM A992 standard,
which provides for a higher yield and tensile strength than A36
steel.
The move to A992 steel raised concerns regarding the
appropriateness of the flange widththickness limits for EBF

Corresponding author. Tel.: +1 612 626 0331; fax: +1 612 626 7750.

E-mail address: tokazaki@umn.edu (T. Okazaki).

c 2006 Elsevier Ltd. All rights reserved.
0143-974X/$ - see front matter
doi:10.1016/j.jcsr.2006.08.004

link sections. A limit of 0.30(E/Fy )1/2 was traditionally

specified for the flange widththickness ratio of EBF links.
This flange slenderness ratio corresponds to 8.5 for A36 steel
(with minimum specified yield strength of Fy = 250 MPa)
and 7.2 for A992 steel (minimum specified Fy = 345 MPa).
A number of rolled wide-flange shapes meet the flange
slenderness limit of 8.5 but do not meet the limit of 7.2,
and thus are disqualified from use as EBF links by the
traditional flange slenderness limit. Meanwhile, the effect of
flange slenderness ratio on link behavior has not been explicitly
addressed in previous research. A secondary concern was the
appropriateness of link overstrength factors used in the capacity
design procedure for EBFs. Link overstrength is defined as
the maximum shear force developed in the link divided by
the plastic shear strength of the link. While the 2005 AISC
Seismic Provisions implicitly assume a link overstrength factor
of 1.5, recent tests on large built-up shear links for use in bridge
applications showed overstrength factors of nearly 2 [3,4].
This has led to concerns that current overstrength factors may
be unconservative, particularly for shapes with heavy flanges,
where shear resistance of the flanges contributes significantly
to overstrength.
An experimental research program was conducted at the
University of Texas at Austin. The initial objective for this
program was to examine flange buckling and overstrength in

752

T. Okazaki, M.D. Engelhardt / Journal of Constructional Steel Research 63 (2007) 751765

Failure of the initial specimens to meet rotation requirements

led to questions concerning the loading protocol for testing EBF
links specified in the 2002 AISC Seismic Provisions [8]. A total
of thirty-seven link specimens were tested during the course of
this investigation. In the rest of this paper, the test specimens
and test procedure is discussed, followed by an overview of test
results. Key observations pertaining to loading history, link web
fracture, and link end welds are discussed. The test data is used
to evaluate the flange slenderness limit, inelastic rotation limit,
and overstrength factor for EBF links prescribed in the 2005
AISC Seismic Provisions [1]. Suggestions on new techniques
for link stiffener design and new detailing for link-to-column
connections are made. While some results from the research
program have been discussed in a previous publication by the
authors [9], these results are also mentioned in this paper for
completeness.
2. Experimental program
2.1. Test setup

Fig. 1. Test setup: (a) energy dissipation mechanism of EBF; (b) schematic
representation of test setup; and (c) details and dimensions.

links constructed of A992 steel. Link specimens with various

sections and lengths were tested for this purpose. As discussed
later, many of these initial specimens failed prematurely due
to fracture of the link web. This type of failure mode was
not typically reported in earlier tests [57], and thus motivated
further testing to investigate the cause of the link web fracture.

A test setup was devised to reproduce the force and

deformation environment imposed on a link in an EBF with
one end of the link attached to a column, as shown in Fig. 1(a).
Fig. 1(b) illustrates that the kinematics of the test setup follows
the energy dissipation mechanism of the EBF. Full details and
dimensions of the test setup are shown in Fig. 1(c). The link
length is indicated in the figures by the letter e. The link
specimens were welded to heavy end plates at each end, as
shown in Fig. 2. The end plates were, in turn, bolted into the
setup, between the vertical column and horizontal beam.

Fig. 2. Details of selected link specimens.

T. Okazaki, M.D. Engelhardt / Journal of Constructional Steel Research 63 (2007) 751765

Table 1
Test section properties
Section
W1019
W1033
W1033 (A)
W1033 (B)
W1033 (C)
W1636
W1068
W1840

Fy (MPa)
Flange
Web

Fu (MPa)
Flange
Web

b f /2t f
Nominal

Actual

367
356
374
379
374
362
319
352

509
507
520
518
518
534
479
499

5.1
9.1
9.1
9.1
9.1
8.1
6.6
5.7

5.2
9.2
9.6
9.2
9.6
7.1
6.6
6.1

405
382
365
402
367
392
404
393

531
503
503
530
503
565
531
527

Note: The tabulated Fy is a static yield stress value, measured with the
test machine cross-heads stationary. The tabulated Fu is a dynamic ultimate
strength, measured with the test machine cross-heads in motion.

2.2. Test specimens

Five different wide-flange shapes were used to construct
the test specimens. All sections were of ASTM A992 steel.
The actual measured yield and ultimate strength values are
listed in Table 1 for samples taken from the edges of the
flanges and from mid-depth of the web. The link sections had
a range of flange widththickness ratios, b f /2t f , to study the
effect of flange slenderness on link behavior. The W1019,
W1068, and W1840 sections satisfied the seismically
compact limit [1] for flanges of 0.30(E/Fy )1/2 (=7.2 for
Fy = 345 MPa). The W1033 was chosen specifically because
its flange slenderness exceeds the seismically compact limit
and is at the compact limit [10] of 0.38(E/Fy )1/2 (=9.2 for
Fy = 345 MPa). The flange slenderness of the W1636, based
on the actual measured dimensions, was substantially smaller
than the nominal slenderness so that this section satisfied the
seismically compact limit. Another consideration for selecting
the test sections was to study the link overstrength for sections
with large ratios of flange to web area. This was based on a
concern that heavy flanges can contribute substantially to the
shear capacity of the section, and therefore generate high levels
of overstrength. The W1068 section was chosen specifically
to investigate this issue. The ratio of the area of one flange to
the area of the web, for the W1068, is approximately 2, which
is near an upper bound for rolled wide-flange shapes normally
used as links.
Table 2 provides a listing of all link test specimens. A
range of link lengths were tested, from short shear yielding
links to long flexural yielding links. Links with a length less
than 1.6M p /V p are dominated by shear yielding, whereas
those longer than 2.6M p /V p are dominated by flexural
yielding [1]. Between these limits, link inelastic response is
heavily influenced by both shear and flexure. The link length
parameter, e/(M p /V p ), listed in Table 2, was evaluated based
on the measured section dimensions and the measured yield
strength values.
Fig. 2 shows schematic views of specimens with various
stiffener details. The stiffener locations are listed in Table 2.
As indicated in the table, the thirty-seven link specimens are
categorized into three groups. All specimens in Groups I and
II were provided with intermediate stiffeners according to the

753

2005 AISC Seismic Provisions. As illustrated in Fig. 2 for

Specimens 4A, 4B, and 4C, stiffeners were provided on only
one side of the web, as permitted by the Provisions for link
sections with a depth less than 635 mm. The stiffeners were
full depth, welded to the web and to both flanges using fillet
welds. The specimens in Group III, which were all identical in
section and length, were provided with varying stiffener details.
Specimens S1 through S3 had full depth one-sided stiffeners,
similar to specimens in Groups I and II. Specimens S4, S6, and
S9 had full depth stiffeners at both sides of the web, welded
to the web and to both flanges. Specimens S5, S7, S8, and
S10 did not meet the stiffener requirements in the Provisions.
Specimens S5 and S8 had stiffeners at both sides of the web,
welded only to the flanges and not to the web. Specimen S7
had one-sided stiffeners welded only to the web and not to
the flanges. The spacing of stiffeners in Specimen S10 did not
meet the stiffener spacing requirement in the Provisions. The
stiffeners for Specimens S6 and S9 were welded to the web
and flanges using a self-shielded flux core arc welding (FCAW)
process with an E70T-6 electrode. In the remaining thirty-five
specimens, the stiffeners were welded to the web and/or flanges
using a shielded metal arc welding (SMAW) process with an
E7018 electrode.
2.3. Loading protocol
Four different cyclic loading protocols, as shown in Fig. 3,
were used in the tests. As indicated in the figure, the four
protocols are referred to in this paper as the old-AISC, revised,
severe, and random loading protocols. Each loading protocol
controls the link rotation angle, , which is computed as the
relative displacement of one end of the link compared to the
other, divided by the link length.
After several initial elastic cycles, the old-AISC loading
protocol (Old) requires increasing the applied link rotation in
increments of 0.01 rad, with two cycles of loading applied
at each increment of rotation. The severe loading protocol
(SEV) was identical to the old-AISC protocol, except that
four cycles of loading, instead of two cycles, were required
at each increment of rotation. This protocol was intended to
promote low cycle fatigue and premature failure of the link
specimen. The Revised Loading Protocol (RLP) requires that,
after completing the loading cycle at a link rotation of 0.05 rad,
the link rotation be increased in increments of 0.02 rad, with
one cycle of loading applied at each increment of rotation.
The old-AISC protocol was specified in the previous, 2002
AISC Seismic Provisions as the loading protocol for testing
EBF links. However, the old-AISC protocol is replaced by the
revised protocol in the current, 2005 AISC Seismic Provisions.
As discussed later, the revised protocol is believed to be more
representative of demands caused by actual earthquake ground
motion than the old-AISC protocol. Except for Specimen
9-RLP, the revised protocol used in this study included two
loading cycles at a link rotation of 0.02 rad, instead of four
loading cycles required in the 2005 AISC Seismic Provisions.
It is believed that the lack of these small amplitude cycles had
limited influence on the overall performance of the specimen.

754

T. Okazaki, M.D. Engelhardt / Journal of Constructional Steel Research 63 (2007) 751765

Table 2
Test specimens
Group

Specimen

Section

Link length
e (mm)

e/(M p /V p )

Intermediate stiffeners

Loading protocol

1A
1B
1C
2
3
4A
4B
4C
5
6A
6B
7
8
9
10
11
12

W1019
W1019
W1019
W1019
W1019
W1033
W1033
W1033
W1033
W1033
W1033
W1033
W1636
W1636
W1068
W1068
W1840

584
584
584
762
1219
584
584
584
930
1219
1219
1854
930
1219
930
1219
584

1.73
1.73
1.73
2.25
3.61
1.04
1.04
1.04
1.65
2.16
2.16
3.29
1.49
1.95
1.25
1.64
1.02

3@146 mm
3@146 mm
3@146 mm
4@152 mm
152 mm from each end
3@146 mma
3@146 mma
3@146 mma
5@156 mm
4@244 mm
4@244 mm
305 mm from each end
6@133 mm
5@203 mm
2@305 mm
3@305 mm
3@146 mm

Old
Old
Old
Old
Old
Old
Old
Old
Old
Old
Old
Old
Old
Old
Old
Old
Old

II

4A-RLP
4C-RLP
8-RLP
9-RLP
10-RLP
11-RLP
12-RLP
12-MON
12-SEV
12-RAN

W1033
W1033
W1636
W1636
W1068
W1068
W1840
W1840
W1840
W1840

584
584
930
1219
930
1219
584
584
584
584

1.04
1.04
1.49
1.95
1.25
1.64
1.02
1.02
1.02
1.02

3@146 mma
3@146 mma
6@133 mm
5@203 mm
2@305 mm
3@305 mm
3@146 mm
3@146 mm
3@146 mm
3@146 mm

RLP
RLP
RLP
RLP
RLP
RLP
RLP
MON
SEV
RAN

III

S1
S2
S3
S4
S5
S6
S7
S8
S9
S10

W1033 (A)
W1033 (B)
W1033 (C)
W1033 (B)
W1033 (B)
W1033 (B)
W1033 (B)
W1033 (B)
W1033 (B)
W1033 (B)

584
584
584
584
584
584
584
584
584
584

1.01
0.99
0.99
0.99
0.99
0.99
0.99
0.99
0.99
0.99

3@146 mma
3@146 mma
3@146 mma
3@146 mma
3@146 mma,b
3@146 mma
3@146 mma,b
3@146 mma,b
3@146 mma
2@195 mma,b

SEV
SEV
SEV
SEV
SEV
SEV
SEV
RLP
RLP
RLP

a See Fig. 2 for stiffener details.

b Violates the stiffener requirements in the 2005 AISC Seismic Provisions.

Finally, the random loading protocol (RAN) was a randomly

generated sequence which imposes large rotations in both
loading directions during early loading cycles.
As indicated in Table 2, the seventeen specimens in Group I
were tested using the old-AISC protocol. The ten specimens in
Group II were tested using various loading histories, including
monotonic loading (MON). The ten specimens in Group III
used either the severe protocol or the revised protocol.
3. Test results
Acceptance criteria for links are defined in the 2005 AISC
Seismic Provisions based on inelastic rotation. The inelastic
rotation, p , is evaluated by removing the contributions of
elastic response from the link rotation, . The Provisions
specify shear yielding links (e 1.6M p /V p ) should be capable
of developing an inelastic rotation of 0.08 rad, whereas flexural
yielding links (e 2.6M p /V p ) should be capable of an

inelastic rotation of 0.02 rad. The required inelastic rotation

of intermediate length links (1.6M p /V p < e < 2.6M p /V p )
is determined by linear interpolation between 0.08 and 0.02
rad. The inelastic rotation capacity of the link specimens was
defined per the 2005 AISC Seismic Provisions, as the maximum
level of inelastic rotation sustained for at least one full cycle
of loading prior to the link shear strength dropping below the
nominal link shear strength. Here, the nominal strength was
evaluated based on the nominal section dimensions and a yield
strength of Fy = 345 MPa.
Table 3 summarizes results for each of the link specimens
tested in this program. The table lists the actual inelastic
rotation achieved by the specimen, along with the inelastic
rotation required by the 2005 AISC Seismic Provisions. Also
listed is a brief description of the controlling failure mode
for each specimen. Detailed descriptions of individual tests
can be found in Arce [11] (for all specimens in Group I
except Specimen 12), Ryu [12] (Specimen 12 and all specimens

T. Okazaki, M.D. Engelhardt / Journal of Constructional Steel Research 63 (2007) 751765

755

Fig. 3. Loading protocols.

Fig. 4. Response of Specimen 4A.

in Group II except Specimen 9-RLP), and Galvez [13] (all

specimens in Group III).
Specimens 1A, 1B, and 6A failed prematurely due to
fractures at the fillet welds connecting the link flanges to the
end plates. These failures are considered an artifact of the
test setup, as the link end connections used for the specimens
were not representative of typical link end connection details
used in actual EBFs. Details of the fracture in the link end
welds and their implications will be discussed later in this
paper. Meanwhile, these specimens will be excluded from the
discussion of link behavior. Excluding Specimens 1A, 1B,
and 6A, there are thirty-four remaining specimens that were
not affected by failures at the link end connections, and can
therefore be considered as providing valid fundamental data on
the behavior of links.

A notable feature of the tests was that specimens with length

e < 1.7M p /V p typically exhibited link web fracture as the
controlling failure mode. The link web fractures initiated at the
top and bottom ends of the link web stiffeners, at the point
of termination of the fillet welds connecting the stiffeners to
the link web. These fractures often propagated in a horizontal
direction, running parallel to the flanges. Ultimately, the growth
of these cracks led to a drastic reduction of the link shear
resistance. Specimen 4A provides an example of a specimen
that failed due to this type of fracture. Fig. 4 shows the
hysteretic response of this specimen and Fig. 5 shows the
specimen after testing. Fractures running across the top and
bottom of the link web are visible in this photo. The link web
fractures limited the inelastic rotation capacity of Specimen 4A
to p = 0.06 rad, falling short of the 0.08 rad required by the

756

T. Okazaki, M.D. Engelhardt / Journal of Constructional Steel Research 63 (2007) 751765

Table 3
Test results
Specimen
1A
1B
1C
2
3
4A
4B
4C
5
6A
6B
7
8
9
10
11
12
4A-RLP
4C-RLP
8-RLP
9-RLP
10-RLP
11-RLP
12-RLP
12-MON
12-SEV
12-RAN
S1
S2
S3
S4
S5
S6
S7
S8
S9
S10

p (rad)
Required

Test

0.072
0.072
0.072
0.042
0.02
0.08
0.08
0.08
0.077
0.046
0.046
0.02
0.08
0.059
0.08
0.078
0.08
0.08
0.08
0.08
0.059
0.08
0.078
0.08
0.08
0.08
0.08
0.08
0.08
0.08
0.08
0.08
0.08
0.08
0.08
0.08
0.08

0.042
0.060
0.081
0.070
0.041
0.061
0.071
0.080
0.067
0.047
0.047
0.037
0.077
0.048
0.073
0.068
0.091
0.12
0.12
0.117
0.058
0.113
0.087
0.119
>0.34
0.072
0.125
0.062
0.061
0.072
0.061
0.071
0.051
0.051
0.122
0.101
0.121

Observed failure mode

Fracture at link end plate connection
Fracture at link end plate connection
Flange and web buckling followed by fracture in web
Flange and web buckling followed by fracture in flange near link end
Flange and web buckling followed by fracture in flange near link end
Fracture of web at stiffener weld
Fracture of web at stiffener weld
Fracture of web at stiffener weld
Fracture of web at stiffener weld
Fracture at link end plate connection
Flange and web buckling followed by fracture in web
Flange, web, and lateral torsional buckling
Flange and web buckling followed by fracture in web at stiffener weld
Flange and web buckling
Fracture of web at stiffener weld
Fracture of web at stiffener weld
Fracture of web at stiffener weld accompanied by web buckling
Fracture of web at stiffener weld
Fracture of web at stiffener weld
Flange and web buckling followed by fracture in web at stiffener weld
Flange and web buckling
Fracture of web at stiffener weld possibly caused by inadequately small fillet weld size
Fracture of web at stiffener weld; link rotation limited by ram stroke
Fracture of web at stiffener weld accompanied by web buckling
Web buckling
Fracture of web at stiffener weld
Web buckling followed by fracture of web at stiffener weld
Fracture of web at stiffener weld
Fracture of web at stiffener weld
Fracture of web at stiffener weld
Fracture of web at stiffener weld
Web buckling followed by fracture of web
Fracture of web at stiffener weld
Fracture of web at stiffener weld promoted by web buckling
Web buckling followed by fracture of web
Fracture of web at stiffener weld
Fracture of web at stiffener weld and web buckling

2005 AISC Seismic Provisions. In fact, the majority of shear

link specimens in Group I, tested with the old-AISC loading
protocol, failed to achieve their required inelastic link rotations
due to this type of fracture. The majority of the twenty-seven
specimens with length e < 1.7M p /V p were controlled by link
web fracture. The only exceptions were Specimen 12-MON,
which was tested with monotonic loading, and Specimens S5
and S8, which did not have the stiffeners welded to the link
web.
Specimens 9 and 9-RLP were identical specimens with
an intermediate length of e = 2.0M p /V p , tested with the
old-AISC protocol and revised protocol, respectively. Both
specimens failed before achieving the rotation requirement,
due to strength degradation associated with severe flange
and web buckling in the end panels. The link web fractures
discussed above were not observed in these specimens. Fig. 6
shows the hysteretic response of Specimen 9-RLP and Fig. 7
shows the specimen after testing. The photo shows substantial
concentration of deformation in the end panels, where the
section was severely distorted due to combined flange and web

Fig. 5. Specimen 4A after testing.

buckling. Significant yielding is visible in the link web panels

besides the end panels. The development of local buckling led
to the gradual strength degradation shown in Fig. 6.
Besides Specimens 9 and 9-RLP, six other specimens had
longer links in the intermediate and flexure yielding range of

T. Okazaki, M.D. Engelhardt / Journal of Constructional Steel Research 63 (2007) 751765

757

Fig. 6. Response of Specimen 9-RLP.

Fig. 7. Specimen 9-RLP after testing.

1.7M p /V p < e 3.6M p /V p . These specimens, all tested

with the old-AISC loading protocol, successfully achieved
their required link rotations, and ultimately failed due to
combinations of severe flange buckling, web buckling, and in
some cases, lateral torsional bucking. The link web fracture
exhibited by short links was not observed in these specimens.
Overall, the behavior of longer links (1.7M p /V p < e
3.6M p /V p ) was very similar to that reported in earlier tests,
for example in Engelhardt and Popov [14].
4. Loading protocol
A large number of test specimens in Group I failed
prematurely, before achieving their required rotation levels.
Most of these specimens met all link design requirements, and
were tested with the old-AISC loading protocol specified in the
2002 AISC Seismic Provisions. After observing these results,
Richards and Uang [15] noted that the typical loading histories
used in shear link tests conducted in the 1970s and 1980s
introduced a significantly smaller number of inelastic loading
cycles compared to the old-AISC loading protocol used for the
tests in Group I. Further, there appeared to be no rational basis
for the old-AISC loading protocol. Consequently, Richards and
Uang [16,17] developed a revised loading protocol for testing

EBF links, which was then used for selected specimens in

Groups II and III. The revised loading protocol was developed
using a methodology similar to that used for moment frame
connection testing, developed under the FEMA/SAC program
by Krawinkler et al. [18].
Seven of the specimens in Group I that failed to meet their
inelastic rotation requirements were duplicated and retested
using the revised loading protocol developed by Richards
and Uang [16,17]. Among the seven specimens, the six
shorter specimens with length e < 1.7M p /V p achieved link
rotations well in excess of the required level. As indicated
by the data in Table 3, these specimens developed inelastic
rotations of 1050% greater than the level required in the 2005
AISC Seismic Provisions. The inelastic rotation developed by
Specimen 11-RLP was limited due to limitations in the stroke
of the loading ram. Nonetheless, this specimen developed an
inelastic rotation exceeding the required level.
However, one specimen tested with the revised loading
protocol, 9-RLP, failed to significantly exceed the required
rotation. After completing a full cycle at p = 0.058 rad,
this intermediate link specimen lost its strength during the
following loading cycle of roughly p = 0.08 rad due
to severe flange and web buckling. Consequently, although
Specimen 9-RLP achieved a 20% greater rotation compared
to Specimen 9, which was tested with the old-AISC loading
protocol, Specimen 9-RLP still failed to meet the required
inelastic rotation of p = 0.059 rad. However, the specimen
retained a shear strength of 120% of its nominal shear strength
at p = 0.058 rad, and the following strength degradation
was rather gradual. As indicated in Fig. 6, the backbone
curve connecting the maximum rotation points at each rotation
increment exceed the nominal shear strength up to p =
0.07 rad, which is beyond the required p = 0.059 rad.
Therefore, since Specimen 9-RLP failed to meet the rigorous
inelastic rotation requirement by such a small margin, and
maintained appreciable strength during rotations beyond the
required inelastic rotation, this specimen is considered as
effectively satisfying the rotation requirements of the 2005
AISC Seismic Provisions.

758

T. Okazaki, M.D. Engelhardt / Journal of Constructional Steel Research 63 (2007) 751765

Fig. 8. Response of Specimen 4A-RLP.

Fig. 9. Specimen 4A-RLP after testing.

To illustrate the effect of loading protocol, Fig. 4 shows

the hysteretic response of Specimen 4A, which was tested
using the old-AISC loading protocol. In comparison, Fig. 8
shows the hysteretic response of Specimen 4A-RLP, which was
nominally identical to Specimen 4A, except that 4A-RLP was
tested using the revised loading protocol. Specimen 4A-RLP
developed an inelastic rotation capacity of p = 0.12 rad,
as compared to p = 0.061 rad for Specimen 4A. Fig. 9 is
a photo of Specimen 4A-RLP after testing, showing the link
web fractures that ultimately caused failure of this specimen.
This failure mode is very similar to that observed in Specimen
4A, shown in Fig. 5. In general, whereas the change from the
old-AISC protocol to the revised protocol led to an increase
in link rotation capacity on the order of 20100%, it did not
significantly change the controlling failure mode for the link.
Links that failed due to fracture of the link web under the oldAISC protocol still failed by fracture of the link web under the
revised protocol.
In order to further investigate the effect of loading history on
link behavior, four replicates of Specimen 12 were fabricated
and tested with different loading histories. In addition to the
revised loading protocol discussed above (Specimen 12-RLP),
three other duplicate specimens were subjected to monotonic
loading (12-MON), the severe loading protocol (12-SEV),

and random loading protocol (12-RAN). Specimen 12-MON

achieved an inelastic rotation larger than 0.34 rad, which is
more than four times the 0.08 rad required in the 2005 AISC
Seismic Provisions. An earlier monotonic loading test by Kasai
and Popov [7] also showed a shear link specimen developing
p = 0.19 rad. These test results demonstrate the ability of EBF
links to withstand very large single excursion deformations,
such as those imposed by large near-fault earthquake ground
motions. The inelastic rotations achieved in the five tests on the
replicates of Specimen 12 were strongly related to the imposed
loading history. The inelastic rotation capacities increased from
Specimen 12-SEV ( p = 0.072 rad), to 12 ( p = 0.091 rad),
to 12-RLP ( p = 0.119 rad), to 12-RAN ( p = 0.125 rad),
and finally, to 12-MON ( p > 0.34 rad, one direction only).
Specimens 12, 12-RLP, 12-SEV, and 12-RAN all failed due to
link web fracture initiating at the termination of stiffener to
link web welds. Specimen 12-MON failed due to severe web
buckling and distortion of the link. Although web fracture was
observed at the stiffener weld terminations, web buckling was
the primary cause of strength loss. Specimen 12-RAN showed
strength degradation due to web buckling before development
of significant link web fracture. These observations suggest that
link web fracture is related to low cycle fatigue effects, and
if very large rotations are imposed at an early loading stage,
significant web buckling can precede the occurrence of link web
fracture.
As illustrated by the discussion above, the loading protocol
used to test link specimens has a very large effect on the
inelastic rotation capacity achieved by the link. Links tested
with the revised loading protocol achieved inelastic rotations
that were from 20 to 100% greater than links tested with the
old-AISC protocol. The average increase in inelastic rotation
using the revised protocol over using the old-AISC protocol
was 47%. Since the loading protocol has such a large effect
on link test results, it is important that loading protocols be
selected that realistically reflect link demands under actual
earthquake loading, as represented by the revised loading
protocol developed by Richards and Uang [16,17].

T. Okazaki, M.D. Engelhardt / Journal of Constructional Steel Research 63 (2007) 751765

5. Link web fracture

The majority of specimens in the length range of e <
1.7M p /V p developed web fractures at the ends of stiffener to
link web welds, prior to the occurrence of any web buckling, or
only after very mild buckling, as exhibited by Specimen 4A (see
Fig. 5). However, this type of link web fracture was not reported
in earlier tests, for example in Hjelmstad and Popov [5], and
Malley and Popov [6]. While fractures of the link web were
reported in these past tests, those fractures occurred only after
the web had undergone severe buckling, with fracture initiation
occurring at locations of large localized bucking deformations.
An exception is a recent test by McDaniel et al. [4] on a
very large built-up link section tested for use in the new east
span of the San FranciscoOakland Bay Bridge. In this test,
a fracture initiated at the termination of a stiffener weld prior
to web buckling, although the fracture propagated diagonally
across the web, compared to the horizontal fracture propagation
observed in Specimen 4A. Analysis of the failure [4] suggested
that the fracture was caused by a stress concentration at the end
of the stiffener weld, because the stiffener was terminated too
close to the flange-to-web groove weld of the built-up section.
The occurrence of pre-bucking web fracture as the
controlling failure mode for shear links was not typically
observed in the extensive link testing programs conducted in
the 1980s, which formed the basis for EBF detailing rules
in the AISC Seismic Provisions. Consequently, links tested in
this current program exhibited fundamentally different failure
mechanisms from many links tested in earlier programs.
Richards and Uang [15] noted three significant differences
between recent and earlier link tests. These differences were
in: (a) cyclic loading history; (b) stiffener details; and (c) link
material. Tests in Groups II and III were conducted to further
study the effect of these three factors, and to further examine
the cause of the link web fracture.
5.1. Loading protocol
An initial concern after completing the tests in Group I
was that the majority of shear yielding links failed due to link
web fracture before achieving the required inelastic rotation
of p = 0.08 rad. As discussed earlier, this concern led to
questions regarding the loading protocol provided in the 2002
AISC Seismic Provisions, and the development of the revised
loading protocol by Richards and Uang [16,17]. Six shear
yielding links were duplicated and retested using the revised
loading protocol. Since all six specimens (4A-RLP, 4C-RLP,
8-RLP, 10-RLP, 11-RLP, and 12-RLP) significantly exceeded
the required p = 0.08 rad, concerns regarding the ability of
shear links to provide adequate inelastic rotation capacity were
largely alleviated.
The shear yielding links tested in this program consistently
failed by link web fracture, regardless of the loading protocol.
The only exceptions were Specimens 12-MON, S5, and S8.
While the monotonically loaded Specimen 12-MON failed due
to web buckling, the web buckling was not evident until the
specimen reached a very large rotation of p = 0.2 rad. The

759

failure modes of Specimens S5 and S8 were associated with

the unique stiffener details used for these specimens, in that
the stiffeners were not welded to the link web. These unique
specimens aside, all other test results indicate that the loading
history has little influence on the link web fracture. That is,
regardless of loading protocol, the shear links tested in this
program consistently exhibited web fracture as the controlling
failure mode.
5.2. Stiffener details
The first specimen in this program to exhibit a horizontal
web fracture initiating at the end of a stiffener weld was
Specimen 4A. Following this failure, and based on the
recommendations by McDaniel et al. [4], the stiffener welds
were terminated at a larger distance from the flange in the
subsequent Specimens 4B and 4C. In going from Specimens
4A to 4B, and then to 4C, the termination of the stiffener
weld was moved progressively further from the flange (see
Fig. 2). In Specimen 4C, the stiffener welds were terminated a
distance of approximately five times the web thickness from the
k-line of the section. The k-line is the location where the web
meets the flangeweb fillet. The test results (see Table 3) show
that larger inelastic rotations were achieved as the stiffener
welds were moved further from the k-line. However, even for
Specimen 4C, which had the stiffener welds terminated quite
a large distance from the k-line, the horizontal fractures still
ultimately developed. Thus, while moving the stiffener welds
further from the k-line was beneficial, it did not eliminate the
occurrence of link web fracture.
Although moving the stiffener weld termination further from
the k-line resulted in higher inelastic rotations prior to link
web fracture, all shear link specimens tested with the revised
loading protocol developed the required inelastic rotation of
0.08 rad. That is, even specimens where the stiffener weld
termination was relatively close to the k-line (Specimen 4ARLP for example) still developed 0.08 rad inelastic rotation.
Therefore, providing a generous distance between the k-line
and the stiffener weld termination, while beneficial, may not
be essential for satisfactory link performance. Nonetheless,
due to the beneficial higher inelastic rotations, terminating
the stiffener welds a generous distance from the k-line is
recommended. Based on this test program, a distance of five
times the link web thickness from the k-line of the link section
to the stiffener weld termination is suggested as a reasonable
basis for design.
As noted by Richards and Uang [15], a large number of
earlier link tests were conducted to investigate stiffener design
criteria, and as such, the majority of these specimens did
not satisfy the current stiffener design criteria. As discussed
above, the link web fracture observed in Groups I and II tests
was not reported previously with the exception of McDaniel
et al. [4], which also followed the current stiffener design
criteria. Therefore, it was speculated that the preclusion of
web buckling shifts the critical failure mode to one controlled
by fracture at the location of high restraints due to low cycle
fatigue. In order to investigate the relation between stiffener

760

T. Okazaki, M.D. Engelhardt / Journal of Constructional Steel Research 63 (2007) 751765

details and link web fracture, and to identify details that

delay link web fracture, specimens in Group III were tested
with various stiffener details. Specimens S4 through S10 were
provided with non-typical stiffener details, some of which
violate requirements in the 2005 AISC Seismic Provisions.
These specimens were constructed of the same material as
Specimen S2, which conformed to the stiffener requirements
in the Provisions. The severe loading protocol was used for
Specimens S4 through S7 in order to promote premature
link web fracture. Further, the specimens had stiffener welds
terminated close to the k-line of the section (see Fig. 2) in order
to draw higher stresses and strains near the k-area, where rolled
shapes can have locally degraded material properties.
Specimen S4 had full depth stiffeners on both sides of the
web, welded to both flanges and to the web. This arrangement
was expected to elevate the local restraint near the k-area
compared to the standard arrangement as in Specimens S2, and
consequently, promote link web fracture. However, Specimen
S4 sustained one more loading cycle than Specimen S2 before
failing by link web fracture. Therefore, two-sided fully welded
stiffeners were no more detrimental to the occurrence of web
fracture than one-sided stiffeners. Specimen S6 had the same
stiffener arrangement as Specimen S4, but used a different
welding process to weld the stiffeners to the link flanges and
web. Specimen S4, which used the shielded metal arc welding
(SMAW) process with an E7018 electrode, achieved greater
rotation and survived four-and-one-half more loading cycles
than Specimen S6, which used the self shielded flux cored arc
welding (FCAW) process with an E70T-6 electrode. The lower
fracture toughness and higher heat input of the FCAW process
may have caused the degraded performance of Specimen S6
compared to Specimen S4.
Specimen S5 was provided with unconventional link
stiffener detailing. This specimen had full depth stiffeners on
both sides of the web, but these stiffeners were welded only
to the flanges and not to the web. Eliminating the stiffener
to link web weld was expected to avoid the occurrence of
fracture in the k-area of the link web. In the absence of stiffener
to link web welds, it was expected that restraint against link
web buckling would be achieved by sandwiching the link
web between stiffeners on opposite sides of the web. Unlike
specimens with fully welded stiffeners, Specimen S5 exhibited
notable web buckling starting from early inelastic loading
cycles. Although no fracture occurred near the k-area of the
section, this specimen ultimately fractured at a location of
concentrated buckling deformation, where the link web was
rubbing against the stiffener. Specimen S5, which had stiffeners
welded to the link flanges only, completed three more loading
cycles and achieved a larger inelastic rotation than Specimen
S2, which was provided with conventional stiffener detailing.
This approach of sandwiching the link web between stiffeners
without applying welds to the web can provide excellent cyclic
performance, and merits further investigation.
Specimen S7 was also provided with unconventional
stiffener detailing. For this specimen, partial depth stiffeners
were provided on only one side of the web, and were welded
only to the web and not to the flanges. This arrangement

was expected to reduce the local restraint near the k-area,

although it was questionable whether the stiffeners would
provide sufficient buckling restraint for the web. Specimen S7
developed a smaller inelastic rotation than Specimen S2, and
failed due to web fracture at the termination of stiffener welds.
Buckling deformation of the web concentrated near the vertical
ends of the stiffeners, and eventually caused fracture of the web.
Therefore, eliminating the connections of the stiffeners to the
flanges was not beneficial for mitigating fracture development
in the web.
Specimens S5 and S6 were duplicated and retested using
the revised loading protocol. These two specimens, designated
respectively as S8 and S9, developed inelastic rotations much
greater than the required 0.08 rad. The failure modes were
similar to the failure modes of Specimens S5 and S6. These
results suggest that the two-sided stiffeners, whether welded or
not welded to the link web, are effective arrangements to control
link rotation capacity.
Specimen S10 was provided with fewer stiffeners (i.e., larger
stiffener spacing) than required in the 2005 AISC Seismic
Provisions. Under the revised loading protocol, this specimen
developed an inelastic rotation much greater than the required
0.08 rad, and failed due to fracture of the link web. Although the
specimen exhibited strength degradation caused by substantial
web buckling, Specimen S10 ultimately failed due to link web
fracture. This test result suggests that the shear link stiffener
spacing requirement in the Provisions is conservative. Further
research may be beneficial to determine if a relaxation in the
stiffener spacing criteria is justified.
Based on the above observations for Specimen S10, the link
web fractures which were observed in the current tests, but not
reported from earlier tests, do not appear to be the result of the
stiffener spacing criteria in the Provisions. Alternative stiffener
designs which are not currently permitted by the Provisions,
such as the sandwiching stiffeners used in Specimens S5 and
S8, may be effective for link detailing. As demonstrated by
the comparison between Specimens S4 and S6, the welding
process used for the stiffener to web weld may have a significant
influence on the link web fracture.
5.3. Material properties
The proximity of the link web fractures in many of the
test specimens to the k-line of the section suggests that
material properties in the k-area may have played a role in
these fractures. The commentary on the 2005 AISC Seismic
Provisions discusses that the steel in the k-area of rolled wideflange shapes (the region where the web meets the flange) can
exhibit high hardness and be prone to fracture.
In order to investigate the effect of material properties on
link behavior, specimens of identical geometry, Specimens S1,
S2, and S3, were constructed from W1033 shapes produced
by three different mills. The three steels are indicated in Table 1
as (A), (B), and (C). Sample ASTM Rockwell B hardness
measures taken along the web-centerlines for the three steel
sections are shown in Fig. 10(a). While all three sections show
notably higher hardness values near the k-area of the section,

T. Okazaki, M.D. Engelhardt / Journal of Constructional Steel Research 63 (2007) 751765

761

performance between the three specimens, S1, S2, and S3. Due
to the higher hardness and lower elongation in their k-areas,
Specimen S1 (Steel (A)) and S3 (Steel (C)) were expected
to exhibit a smaller rotation capacity compared to Specimen
S2 (Steel (B)). However, the overall performance of the three
specimens was quite similar, with Specimen S3 completing
three more loading cycles than Specimens S1 and S2. All three
specimens failed by fracture of the link web initiating at the
stiffener weld terminations. While these test results indicate that
the steel material can influence link performance, the results
do not clarify the correlation between the reduced material
ductility in the k-area and the occurrence of link web fracture.
It is possible that the difference in k-area material properties
between the three steels, which were all A992 steel, was
insufficient to clearly highlight the influence of k-area material
properties.
A rather large number of tests in this program indicate that
the loading history and stiffener details do not change the failure
mode of shear links from that controlled by link web fracture
to a different failure mode. However, the relation between the
material properties and the link web fracture is less clear. It is
suggested that further studies be conducted to clarify the effect
of material properties, specifically the properties in the k-area,
on link web fracture.
6. Link end welds

Fig. 10. Material properties measured for W1033 sections: (a) hardness
measures along web centerline of link section; and (b) tension coupon test
results for Steel (C).

Steel (A) shows the highest peak hardness value, while Steel
(B) shows somewhat lower hardness values. The peak value
is seen at a distance of 2025 mm from the outer face of
the flange, at the k-line of the section. Tension coupons were
taken from the location of the web indicated in Fig. 10(a). For
Steels (A) and (C), the coupons taken from the k-area showed
2535% higher tensile strength and a two-thirds reduction in
elongation compared to the coupon taken from the mid-depth
of the web. For Steel (B), the coupons taken from the k-area
showed a 5% higher tensile strength and one-third reduction in
elongation compared to the coupon taken from the mid-depth
of the web. Fig. 10(b) shows the tensile coupon test results
for Steel (C), comparing three coupons taken from the edge
of the flange, mid-depth of the web, and k-area of the section.
The figure illustrates the significantly higher tensile strength
and reduced elongation of the coupon taken from the k-area.
Elevated hardness values, elevated tensile strength, and reduced
ductility is characteristic of the k-area of roller straightened
shapes [19], and was observed in all sections used for the link
specimens in this study with the exception of the W1019.
For specimens in Group III, the stiffener weld was
terminated within the region of elevated hardness values, as
indicated in Fig. 10(a), so that the degraded material properties
would influence their performance. The notable difference in
the k-area material properties was expected to cause varying

The link specimens tested in this program were welded to

heavy end plates, which in turn were bolted into the test setup
shown in Fig. 1. Studying the behavior of connections between
the link and surrounding members was not an objective of
this research program. Consequently, the end plate connection
details used in these tests were not intended to represent
realistic link end conditions. Rather, the end plate detail was
devised to preclude failure at the link ends, and thereby to
permit study of link behavior. Nonetheless, the results of this
test program provide some useful insights into potentially
effective link-to-column connection details.
As discussed earlier, Specimens 1A, 1B, and 6A failed
prematurely due to fracture of a fillet weld connecting the link
flange to the end plate. The fracture of the end plate welds
occurred either in the throat of the weld, or in the link flange
base metal near the weldbase metal interface. Similar fractures
were reported in tests by Kasai and Popov [7] and Ramadan
and Ghobarah [20]. In order to avoid these fractures, in later
specimens, the leg size of the fillet welds was increased to oneand-one-half times the link flange or web thickness (see Fig. 2).
Further, weld tabs were used at the edges of the flanges, as
shown in Fig. 11, to avoid introducing undercuts or weld defects
at these edges. With the exception of Specimen 1A, which used
the FCAW process with an E70T-6 electrode, the fillet welds
were made using the SMAW process with an E7018 electrode.
Instead of using fillet welds, the 20 mm-thick flanges of the
W1068 links were connected to the end plates with partial
penetration groove welds (groove depth of 14 mm), reinforced
at the root side of the weld by fillet welds with a leg size equal
to the flange thickness.

762

T. Okazaki, M.D. Engelhardt / Journal of Constructional Steel Research 63 (2007) 751765

Fig. 11. Typical end plate weld.

The improved end plate welds shown in Fig. 11 were used

at both ends of thirty-one link specimens. No damage in the
end plate welds was observed in any of these specimens.
Therefore, although the thickness of the end plates of 50 mm
(three to five times the thickness of the link flange) may be
unrealistic for practical application, the end plate connections
used in these tests are promising for application to link-tocolumn connections. Research is continuing by the authors to
develop practical link-to-column connection details.
7. Design implications
7.1. Link flange slenderness limit
A basic objective of this research program was to study the
effect of flange slenderness on link behavior. More specifically,
the objective was to determine if the flange slenderness limit,
b f /2t f , for link flanges can be increased from the seismically
compact limit of 0.30(E/Fy )1/2 to the compact limit of
0.38(E/Fy )1/2 . In this test program, the flange slenderness
of the W1033 was very close to 0.38(E/Fy )1/2 for Fy =
345 MPa. Specimens with W1033 sections were tested
over a range of lengths varying from 1.0 to 3.6M p /V p ,
covering a wide range of link behaviors ranging from shear to
flexural dominated response. The longer W1033 specimens
(Specimens 6B and 7) performed very well. These specimens
exhibited flange buckling, but significantly exceeded their
required rotation levels before link strength dropped below
the defined failure threshold. The shorter W1033 specimens
tested using the old-AISC loading protocol (Specimens 4A
to 4C and 5) did not achieve their required rotations due
to premature web fractures. However, when retested using
the more realistic revised loading protocol, short W1033
specimens (Specimens 4A-RLP and 4C-RLP) significantly
exceeded their required rotation. Similar results were obtained
from Specimens S1 through S10. Therefore, data from the
W1033 test specimens suggest that the compact limit for
flange slenderness is adequate for links of all practical lengths.
Companion finite element simulations of EBF links by Richards

and Uang [15], calibrated to the results of these tests, and

extended to a wider range of link parameters, also support the
use of the compact limit for all link lengths.
The specimens constructed with W1636 sections also
provide useful data on flange slenderness effects. Based on
nominal section dimensions, the flange slenderness of the
W1636 falls between the seismically compact limit and the
compact limit. However, based on measured dimensions, the
actual flange slenderness was smaller than the nominal value,
and fell just within the seismically compact limit. The failure
mode of Specimens 8 and 8-RLP, which were W1636 links
with a length of e = 1.5M p /V p , was unique compared
with other shear yielding links. The specimens developed
significant flange and web buckling at both ends of the link,
and ultimately failed by web fracture at the link end panels.
Although the flange and web buckling did not directly cause
strength degradation, it appeared that the severe web buckling
triggered rapid growth of the web crack. Specimens 9 and
9-RLP, which were W1636 links with a length of e =
2.0M p /V p , failed due to strength degradation associated with
combined flange and web buckling in the link end panels.
Between the two W1636 links tested using the revised
loading protocol, Specimen 8-RLP significantly exceeded the
required rotation, while Specimen 9-RLP only barely met the
rotation requirement, as discussed earlier. The tendency of
local buckling in the W1636 sections suggests caution in
permitting the compact limit of 0.38(E/Fy )1/2 for the flanges
of longer flexure dominated links. The strength degradation
in the W1636 links appeared to result from flangeweb
interaction, and further studies of such interaction are needed.
There is strong and consistent evidence from the results of
this testing program, results from analytical studies [15], as well
as results from previous tests [7], to support the less stringent
limit of 0.30(E/Fy )1/2 for shear yielding links. For longer
links (e > 1.6M p /V p ), the evidence on flange slenderness
effects on link rotation capacity is not as clear. A number of
longer link specimens with a flange slenderness at the limit
of 0.38(E/Fy )1/2 provided excellent performance, achieving
inelastic rotations well beyond the required levels. However,
rather short specimens (e = 1.5 and 2.0M p /V p ) constructed
with the W1636 section showed a notable tendency for flange
buckling. Since the flange slenderness of the W1636 was
within the limit of 0.30(E/Fy )1/2 , it is unclear if the flange
slenderness limit for links dominated by flexure could be
relaxed to 0.38(E/Fy )1/2 . Based on this research and others,
the flange slenderness limit for shear links (e 1.6M p /V p )
has been relaxed to 0.38(E/Fy )1/2 in the 2005 AISC Seismic
Provisions.
7.2. Inelastic rotation limit
The loading protocol used for testing has a very significant
effect on the performance of link specimens. This can be seen
in Fig. 12(a), which compares the inelastic rotation measured
for all twenty-four shear yielding links tested in this program.
Although the data is skewed by the large number of W1033
links (fifteen specimens) and W1840 links (five specimens),

T. Okazaki, M.D. Engelhardt / Journal of Constructional Steel Research 63 (2007) 751765

763

Fig. 12. Performance of shear yielding links: (a) inelastic rotation; and (b) overstrength factor.

each of identical length, the figure indicates the inelastic

rotation is significantly affected by the loading protocol used
for testing. The average inelastic rotation was 0.075 rad for the
six specimens tested with the old-AISC protocol, 0.117 rad for
the eight specimens tested with the revised protocol, and 0.062
rad for the eight specimens tested with the severe protocol.
While the influence of link section geometry, material, and
stiffener details may also be recognized from the figure, the
influence of loading protocol is dominant over these secondary
factors. The data in Fig. 12(a) emphasize the importance of
choosing a loading protocol that realistically reflects expected
earthquake demands, such as the revised loading protocol for
EBF links developed by Richards and Uang [16,17]. This
revised loading protocol has been adopted by the 2005 AISC
Seismic Provisions [1].
All shear yielding links tested in this program with the
revised loading protocol consistently exceeded the required
inelastic rotation of p = 0.08 rad. With the exception
of Specimen 9, the intermediate and flexural yielding links,
tested with the old-AISC protocol, met their inelastic rotation
requirements. Specimen 9-RLP, which was a duplicate of
Specimen 9, tested with the revised protocol, exceeded its
required rotation. Analysis by Richards and Uang [17] suggests
that while the old-AISC protocol was overly conservative for

shear yielding links, it was adequate for flexure yielding links.

Therefore, data from the current project suggest that links
designed according to the 2005 AISC Seismic Provisions would
reliably develop the prescribed inelastic rotation.
7.3. Link overstrength
The link overstrength values evaluated for each specimen
are presented in Table 4. This table lists the ratio Vmax /Vn ,
where Vmax is the largest shear force measured in a test. Vn
is the plastic strength of the link, and was calculated per
the 2005 AISC Seismic Provisions as the smaller of V p or
2M p /e, where V p and M p were computed using the actual
measured dimensions and actual measured yield strengths of
the test sections. As indicated in Figs. 4, 6 and 8, the measured
value of Vn was typically equal to or somewhat larger than its
nominal value (based on nominal dimensions and nominal yield
strength).
The overstrength tended to be greater for shorter links of
lengths between 1.0 and 1.7M p /V p , compared to longer links
with a length greater than 1.7M p /V p . The average overstrength
for these shorter link specimens was 1.41, with a variation from
1.25 to 1.62. The average overstrength for the longer specimens
(e > 1.7M p /V p ) was 1.20, with a variation from 1.05 to 1.27.

764

T. Okazaki, M.D. Engelhardt / Journal of Constructional Steel Research 63 (2007) 751765

Table 4
Specimen overstrength
Specimen

Vmax /V p

Specimen

Vmax /V p

1A
1B
1C
2
3
4A
4B
4C
5
6A
6B
7
8
9
10
11
12

1.20
1.20
1.23
1.24
1.26
1.40
1.42
1.41
1.34
1.21
1.21
1.27
1.35
1.11
1.44
1.42
1.40

4A-RLP
4C-RLP
8-RLP
9-RLP
10-RLP
11-RLP
12-RLP
12-MON
12-SEV
12-RAN
S1
S2
S3
S4
S5
S6
S7
S8
S9
S10

1.45
1.47
1.37
1.05
1.47
1.42
1.44
1.59
1.36
1.62
1.49
1.29
1.56
1.37
1.25
1.32
1.28
1.27
1.43
1.36

Fig. 12(b) compares the overstrength factor measured for

all shear yielding links tested in this program. The average
link overstrength was 1.41 for the specimens tested with
the old-AISC protocol, 1.41 for the specimens tested with
the revised protocol, and 1.37 for the specimens tested with
the severe protocol. Specimens 12 and 12-RLP, 12-SEV,
12-MON, and 12-RAN were subjected to various different
loading histories. Specimen 12-SEV, which was subjected to
a large number of inelastic loading cycles before failing at a
rotation of p = 0.072 rad, developed an overstrength of
1.36, whereas the monotonically loaded Specimen 12-MON
achieved an overstrength of 1.59 at a rotation of roughly p =
0.21 rad. Comparing the seven identical specimens (five of
which are shear yielding links) tested with both the old-AISC
and revised loading protocol, the overstrength was 1.36 for the
specimens tested with the old-AISC protocol, and 1.38 for the
specimens tested with the revised protocol. The data suggest
that the loading protocol has a more limited influence on link
overstrength than on link rotation.
Specimens 10, 11, 10-RLP, and 11-RLP were made of the
W1068 section, which had a high ratio of flange to web area.
These specimens were expected to develop shear resistance in
the flanges, and therefore, greater overstrength compared to
the other specimens. However, these specimens did not show
unusually large values of link overstrength compared to other
shear link specimens.
A notably large variation in overstrength values is seen
between Specimens S1, S2, and S3. These three specimens
were W1033 links of identical length, tested with the severe
loading protocol, but constructed from different steels (see
Table 1 and Fig. 10). The higher overstrength values of
Specimens S1 and S3 compared to Specimen S2 may be
partly due to the significantly higher tensile strength measured
in the k-area of the section, which was not accounted for
in evaluating the nominal shear strength. However, since the

k-area would represent a smaller fraction of the web depth,

the higher strength in the k-area should have a more limited
influence on the overstrength factor of deeper sections. The data
for Specimens S2 and S4 through S7, which were constructed
from the same steel but provided with different stiffener details,
suggest that the stiffener details have little influence on the
overstrength factor.
While the test data suggest smaller overstrength factors
for longer links in the length range of e > 1.7M p /V p , the
W1636 links showed especially low overstrength values. The
overstrength values for Specimens 9 and 9-RLP, both with a
length of e = 2.0M p /V p , were 1.11 and 1.05, respectively.
Combined flange and web buckling caused drastic strength
degradation in these specimens. In fact, links with a length
near e = 2.0M p /V p are expected to experience significant
shear-flexure interaction [14], and thus, the inelastic strength
as defined by the 2005 AISC Seismic Provisions is likely
to be conservative for these links. The occurrence of local
buckling and overestimation of the inelastic strength were the
likely reasons for the particularly low overstrength values for
Specimens 9 and 9-RLP.
8. Conclusions
This paper summarized an experimental program on the
cyclic loading behavior of EBF links made of ASTM A992
steel. The test results provide data pertaining to a wide range of
design issues for EBFs including the flange slenderness limits,
overstrength factors, and stiffener design for links.
Results of this test program clearly show that the loading
protocol used to test EBF links has a very large effect on
the inelastic rotation achieved by the links. Since the loading
protocol has such a large effect on link test results, it is
important that loading protocols realistically reflect demands
caused by actual earthquake loading. Such a loading protocol
for EBF links was recently developed by Richards and
Uang [16,17], and has been adopted by the 2005 AISC Seismic
Provisions.
A number of shear yielding links tested in this program
failed due to fracture of the link web. These fractures initiated at
terminations of stiffener to link web fillet welds and ultimately
caused rapid strength degradation. This type of link web
fracture has not typically been observed in earlier link tests
reported in the literature. Observations from this program
suggest that the loading history and stiffener arrangement have
limited influence on the link web fracture. Further investigation
is recommended to study the effect of material properties in the
k-area on the occurrence of the link web fracture.
Link web fracture can be delayed and link rotation capacity
can be enhanced by altering the stiffener details. One method is
to increase the distance from the k-line of the rolled link section
to the termination of the stiffener to link web fillet weld. Based
on the test results, it is recommended that stiffener welds be
terminated a distance of at least five times the web thickness
from the k-line of the link section. Another method to delay web
fracture is to restrain both sides of the link web using stiffeners
without placing welds directly to the web. The stiffeners are

T. Okazaki, M.D. Engelhardt / Journal of Constructional Steel Research 63 (2007) 751765

welded only to the flanges and not to the web. This technique,
while promising, requires further investigation.
Test data from this program indicate that the flange
slenderness limit of shear yielding links (e 1.6M p /V p ) can
be relaxed to the compact limit of 0.38(E/Fy )1/2 . For longer
links (e > 1.6M p /V p ), it is recommended that the flange
slenderness limit be maintained at the seismically compact
limit of 0.30(E/Fy )1/2 , pending further study of the effects of
shear-flexure interaction on local buckling. All link specimens
conforming to these flange slenderness limits were capable
of achieving the inelastic rotations required in the 2005 AISC
Seismic Provisions.
The ASTM A992 rolled wide-flange links tested in this
program exhibited overstrength factors ranging from 1.05 to
1.62, with an overall average of 1.35. Sections with high
ratios of flange to web areas did not exhibit unusually high
overstrength factors, at least within the range of flange to
web area ratios typical of rolled wide-flange shapes. The
overstrength factor of 1.5, which forms the basis for the
capacity design procedure in the 2005 AISC Seismic Provisions,
appears reasonable for links constructed of typical rolled
shapes. However, based on experimental and analytical results
reported by others, a higher overstrength factor may be
appropriate for short links constructed of built-up shapes with
heavy flanges [15].
As a closing remark, the large number of tests conducted in
this research program suggests that EBF links constructed of
ASTM A992 steel and designed according to the 2005 AISC
Seismic Provisions perform well, and meet the performance
requirements of the 2005 AISC Seismic Provisions.
Acknowledgements
The writers gratefully acknowledge primary funding
provided for this project by the American Institute of Steel
Construction (AISC) and the National Science Foundation
(Grant No. CMS-0000031). The first author expresses gratitude
for sponsorship provided by the Twenty-First Century Center
of Excellence Program awarded to the Tokyo Institute of
Technology, Japan. The tests discussed herein were conducted
as Masters thesis work by former students at the University
of Texas at Austin, Gabriela Arce, Han-Choul Ryu, and Pedro
Galvez. The writers would like to particularly thank Tom
Schlafly of AISC for his support and assistance throughout
this project. The writers thank Chia-Ming Uang, Paul Richards,
James Malley, Subhash Goel, and Tom Sabol for their
assistance and advice on this study.
References
[1] American Institute of Steel Construction, Inc. (AISC) . Seismic provisions
for structural steel buildings. Standard ANSI/AISC 341-05. Chicago (IL,
USA): AISC; 2005.

765

[2] Popov EP, Engelhardt MD. Seismic eccentrically braced frames. Journal
of Constructional Steel Research 1988;10:32154.
[3] Itani AM, El-Fass S, Douglas BM. Behavior of built-up shear links under
large cyclic displacement. Engineering Journal, American Institute of
Steel Construction 2003;40(4):22134.
[4] McDaniel CC, Uang C-M, Seible F. Cyclic testing of built-up steel shear
links for the new bay bridge. Journal of Structural Engineering, American
Society of Civil Engineers 2003;129(6):8019.
[5] Hjelmstad KD, Popov EP. Seismic behavior of active beam link in
eccentrically braced frames. Report No. UCB/EERC-83/15. Berkeley
(Richmond, CA, USA): Earthquake Engineering Research Center,
University of California; 1983.
[6] Malley JO, Popov EP. Design considerations for shear links in
eccentrically braced frames. Report No. UCB/EERC-83/24. Berkeley
(Richmond, CA, USA): Earthquake Engineering Research Center,
University of California; 1983.
[7] Kasai K, Popov EP. A study of seismically resistant eccentrically braced
frames. Report No. UCB/EERC-86/01. Berkeley (Richmond, CA, USA):
Earthquake Engineering Research Center, University of California; 1986.
[8] American Institute of Steel Construction, Inc. (AISC) . Seismic provisions
for structural steel buildings. Standard ANSI/AISC 341-02. Chicago (IL,
USA): AISC; 2002.
[9] Okazaki T, Arce G, Ryu H-C, Engelhardt MD. Experimental study of
local buckling, overstrength, and fracture of links in eccentrically braced
frames. Journal of Structural Engineering, American Society of Civil
Engineering 2005;131(10):152635.
[10] American Institute of Steel Construction, Inc. (AISC) . Specification for
structural steel buildings. Standard ANSI/AISC 360-05. Chicago (IL,
USA): AISC; 2005.
[11] Arce G. Impact of higher strength steels on local buckling and
overstrength in eccentrically braced frames. Masters thesis. Austin (TX,
USA): Department of Civil Engineering, University of Texas at Austin;
2002.
[12] Ryu H-C. Effects of loading history on the behavior of links in seismicresistant eccentrically braced frames. Masters thesis. Austin (TX, USA):
Department of Civil Engineering, University of Texas at Austin; 2005.
[13] Galvez P. Investigation of factors affecting web fractures in shear links.
Masters thesis. Austin (TX, USA): Department of Civil Engineering,
University of Texas at Austin; 2004.
[14] Engelhardt MD, Popov EP. Behavior of long links in eccentrically braced
frames. Report No. UCB/EERC-89/01. Berkeley (Richmond, CA, USA):
Earthquake Engineering Research Center, University of California; 1989.
[15] Richards P, Uang C-M. Evaluation of rotation capacity and overstrength
of links in eccentrically braced frames (phase 1). Report No. SSRP2002/18. La Jolla (CA, USA): Department of Structural Engineering,
University of California at San Diego; 2002.
[16] Richards P, Uang C-M. Development of testing protocol for short links
in eccentrically braced frames. Report No. SSRP-2003/08. La Jolla (CA,
USA): Department of Structural Engineering, University of California at
San Diego; 2003.
[17] Richards P. Cyclic stability and capacity design of steel eccentrically
braced frames. Ph.D. dissertation. La Jolla (CA, USA): University of
California, San Diego; 2004.
[18] Krawinkler H, Gupta A, Medina R, Luco N. Loading histories for seismic
performance testing of SMRF components and assemblies. Report No.
SAC/BD-00/10. Sacramento (CA, USA): SAC Joint Venture; 2000.
[19] Miller KR, Frank K. Study of the material properties of the webflange
intersection of rolled shapes. Report No. SAC/BD-99/08. Sacramento
(CA, USA): SAC Join Venture; 1999.
[20] Ramadan T, Ghobarah A. Behavior of bolted linkcolumn joints in
eccentrically braced frames. Canadian Journal of Civil Engineering 1995;
22:74554.